FALCON: A Fast and Low-Power Current-Mode Near-Sensor-Computing Architecture for Real-Time Edge Visual Processing

Liang Zhang; Jing Kou; Jinyao Mi; Yang Liu; Junda Zhao; Junzhan Liu; Wang Kang

doi:10.23919/DATE69613.2026.11539336

FALCON: A Fast and Low-Power Current-Mode Near-Sensor-Computing Architecture for Real-Time Edge Visual Processing

Liang Zhang
, Jing Kou
, Jinyao Mi
, Yang Liu
, Junda Zhao
, Junzhan Liu^*
, Wang Kang^*

^*Corresponding author for this work

School of Integrated Circuit Science and Engineering

Beihang University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Near-sensor computing (NSC) has emerged as a promising paradigm for edge visual processing and data compression, to mitigate data transmission and computing overheads at IoT nodes. However, existing NSC still suffers from limited precision, reduced frame rate and low energy efficiency under complex DNN tasks due to inefficient analog memory, exponential computation overheads and considerable ADC burden. This paper introduces FALCON, a novel current-mode (CM) NSC architecture featuring in-current-register-processing (ICRP) unit and two-step multiply-and-accumulate (TS-MAC) for high-precision and low-latency feature extraction. Additionally, a reconfigurable ADC with embedded ReLU and pooling functionality is employed to improve ADC overhead and compression ratio. Implemented under a 55nm CIS process, FALCON achieves 12.92 TOPS/W with 7-bit weight precision and supports a frame rate of 3096 fps under 8 filters, with an iFOM of 10.1 pJ/pix•fps.

Original language	English
Title of host publication	2026 Design, Automation and Test in Europe Conference, DATE 2026 - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9783982674117
DOIs	https://doi.org/10.23919/DATE69613.2026.11539336
State	Published - 2026
Event	2026 Design, Automation and Test in Europe Conference, DATE 2026 - Verona, Italy Duration: 20 Apr 2026 → 22 Apr 2026

Publication series

Name	Proceedings -Design, Automation and Test in Europe, DATE
ISSN (Print)	1530-1591

Conference

Conference	2026 Design, Automation and Test in Europe Conference, DATE 2026
Country/Territory	Italy
City	Verona
Period	20/04/26 → 22/04/26

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 7 Affordable and Clean Energy

Keywords

analog-to-digital converter
CMOS image sensor
convolutional neural network
near-sensor-computing

Access to Document

10.23919/DATE69613.2026.11539336

Cite this

Zhang, L., Kou, J., Mi, J., Liu, Y., Zhao, J., Liu, J., & Kang, W. (2026). FALCON: A Fast and Low-Power Current-Mode Near-Sensor-Computing Architecture for Real-Time Edge Visual Processing. In 2026 Design, Automation and Test in Europe Conference, DATE 2026 - Proceedings (Proceedings -Design, Automation and Test in Europe, DATE). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.23919/DATE69613.2026.11539336

@inproceedings{c5142d08a2164b31a06fe1404b6277d3,

title = "FALCON: A Fast and Low-Power Current-Mode Near-Sensor-Computing Architecture for Real-Time Edge Visual Processing",

abstract = "Near-sensor computing (NSC) has emerged as a promising paradigm for edge visual processing and data compression, to mitigate data transmission and computing overheads at IoT nodes. However, existing NSC still suffers from limited precision, reduced frame rate and low energy efficiency under complex DNN tasks due to inefficient analog memory, exponential computation overheads and considerable ADC burden. This paper introduces FALCON, a novel current-mode (CM) NSC architecture featuring in-current-register-processing (ICRP) unit and two-step multiply-and-accumulate (TS-MAC) for high-precision and low-latency feature extraction. Additionally, a reconfigurable ADC with embedded ReLU and pooling functionality is employed to improve ADC overhead and compression ratio. Implemented under a 55nm CIS process, FALCON achieves 12.92 TOPS/W with 7-bit weight precision and supports a frame rate of 3096 fps under 8 filters, with an iFOM of 10.1 pJ/pix•fps.",

keywords = "analog-to-digital converter, CMOS image sensor, convolutional neural network, near-sensor-computing",

author = "Liang Zhang and Jing Kou and Jinyao Mi and Yang Liu and Junda Zhao and Junzhan Liu and Wang Kang",

note = "Publisher Copyright: {\textcopyright} 2026 EDAA.; 2026 Design, Automation and Test in Europe Conference, DATE 2026 ; Conference date: 20-04-2026 Through 22-04-2026",

year = "2026",

doi = "10.23919/DATE69613.2026.11539336",

language = "英语",

series = "Proceedings -Design, Automation and Test in Europe, DATE",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

booktitle = "2026 Design, Automation and Test in Europe Conference, DATE 2026 - Proceedings",

address = "美国",

}

Zhang, L, Kou, J, Mi, J, Liu, Y, Zhao, J, Liu, J & Kang, W 2026, FALCON: A Fast and Low-Power Current-Mode Near-Sensor-Computing Architecture for Real-Time Edge Visual Processing. in 2026 Design, Automation and Test in Europe Conference, DATE 2026 - Proceedings. Proceedings -Design, Automation and Test in Europe, DATE, Institute of Electrical and Electronics Engineers Inc., 2026 Design, Automation and Test in Europe Conference, DATE 2026, Verona, Italy, 20/04/26. https://doi.org/10.23919/DATE69613.2026.11539336

FALCON: A Fast and Low-Power Current-Mode Near-Sensor-Computing Architecture for Real-Time Edge Visual Processing. / Zhang, Liang; Kou, Jing; Mi, Jinyao et al.
2026 Design, Automation and Test in Europe Conference, DATE 2026 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2026. (Proceedings -Design, Automation and Test in Europe, DATE).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - FALCON

T2 - 2026 Design, Automation and Test in Europe Conference, DATE 2026

AU - Zhang, Liang

AU - Kou, Jing

AU - Mi, Jinyao

AU - Liu, Yang

AU - Zhao, Junda

AU - Liu, Junzhan

AU - Kang, Wang

PY - 2026

Y1 - 2026

N2 - Near-sensor computing (NSC) has emerged as a promising paradigm for edge visual processing and data compression, to mitigate data transmission and computing overheads at IoT nodes. However, existing NSC still suffers from limited precision, reduced frame rate and low energy efficiency under complex DNN tasks due to inefficient analog memory, exponential computation overheads and considerable ADC burden. This paper introduces FALCON, a novel current-mode (CM) NSC architecture featuring in-current-register-processing (ICRP) unit and two-step multiply-and-accumulate (TS-MAC) for high-precision and low-latency feature extraction. Additionally, a reconfigurable ADC with embedded ReLU and pooling functionality is employed to improve ADC overhead and compression ratio. Implemented under a 55nm CIS process, FALCON achieves 12.92 TOPS/W with 7-bit weight precision and supports a frame rate of 3096 fps under 8 filters, with an iFOM of 10.1 pJ/pix•fps.

AB - Near-sensor computing (NSC) has emerged as a promising paradigm for edge visual processing and data compression, to mitigate data transmission and computing overheads at IoT nodes. However, existing NSC still suffers from limited precision, reduced frame rate and low energy efficiency under complex DNN tasks due to inefficient analog memory, exponential computation overheads and considerable ADC burden. This paper introduces FALCON, a novel current-mode (CM) NSC architecture featuring in-current-register-processing (ICRP) unit and two-step multiply-and-accumulate (TS-MAC) for high-precision and low-latency feature extraction. Additionally, a reconfigurable ADC with embedded ReLU and pooling functionality is employed to improve ADC overhead and compression ratio. Implemented under a 55nm CIS process, FALCON achieves 12.92 TOPS/W with 7-bit weight precision and supports a frame rate of 3096 fps under 8 filters, with an iFOM of 10.1 pJ/pix•fps.

KW - analog-to-digital converter

KW - CMOS image sensor

KW - convolutional neural network

KW - near-sensor-computing

UR - https://www.scopus.com/pages/publications/105041977022

U2 - 10.23919/DATE69613.2026.11539336

DO - 10.23919/DATE69613.2026.11539336

M3 - 会议稿件

AN - SCOPUS:105041977022

T3 - Proceedings -Design, Automation and Test in Europe, DATE

BT - 2026 Design, Automation and Test in Europe Conference, DATE 2026 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 20 April 2026 through 22 April 2026

ER -

Zhang L, Kou J, Mi J, Liu Y, Zhao J, Liu J et al. FALCON: A Fast and Low-Power Current-Mode Near-Sensor-Computing Architecture for Real-Time Edge Visual Processing. In 2026 Design, Automation and Test in Europe Conference, DATE 2026 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2026. (Proceedings -Design, Automation and Test in Europe, DATE). doi: 10.23919/DATE69613.2026.11539336

FALCON: A Fast and Low-Power Current-Mode Near-Sensor-Computing Architecture for Real-Time Edge Visual Processing

Abstract

Publication series

Conference

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this